Post-shift accepted-trajectory risk estimation on Multi-seed RAG tool mean
0.5Risk Estimation (Post-shift Accepted Trajectory)ToolChain-CRC
Evaluation Results
| Method | Links | |
|---|---|---|
| ToolChain-CRCCalibration Strategy=Trajectory calibration2026.06 | 0.5 | |
| Fixed trajectoryCalibration Strategy=Fixed trajectory2026.06 | 0.6 | |
| Prompt/retrieval-onlyCalibration Strategy=Prompt/retrieval-only2026.06 | 5.1 | |
| Final-onlyCalibration Strategy=Final-answer-only2026.06 | 10.2 |